Ensemble Learning based EEG Classification – Investigating the Effects of Combined Yoga and Rajyog Meditation

Abstract

The ability to detect and prevent mental health deterioration has been one of the major achievements of digital psychiatry using artificial intelligence and machine learning. The aim of this paper is to address the issue of preventing the mental health disorders of young generation by developing a system to predict the changes in an individual's states of psychological health. Pre-and post-yoga and Rajyoga meditation states were analyzed for classification of data. Also, the paper investigates if bidirectional long-short-term memory BiLSTM-based ensemble models outperform the CNN-based models in prediction modeling. The EEG data was collected from 69 students for pre- and post-intervention. To determine an objective marker for yoga and meditation, collected data were analyzed using spectrum analysis, and classification. The post meditation group exhibited highest band powers and wavelet coefficients, indicating the differences in meditation and control conditions. Additionally, in this study, an ensemble model classifier has been developed utilizing EEG data that was more accurate (82%) than other models at differentiating between meditation and control situations. To the best of the knowledge of the authors, this is the first research to apply ensemble model-based classifiers to distinguish between states of meditation and non-meditation. The performance of BiLSTM-DT was the highest among all other models in terms of precision, recall, f-measure, and accuracy. Therefore, the BiLSTM-DT ensemble model is a viable objective marker for psychological health states.